Ckemicd finginwing Science. 1976, Vol. 31. pp. 569-577. Pergamon Pres.

Prinied in Great Brilah

NEW SCALE-UP AND DESIGN METHOD FOR STIRRER

AGITATED BATCH MIXING VESSELS

SOON J. KHANG and OCTAVE LEVENSPIEL

Department of Chemical Engineering, Oregon State University, Corvallis, OR 97331,U.S.A.

(Received 10June 1975;accepted in reuisedform 17 October 1975)

Abstract-Residence time distrihution theory and probability conditions predict that the infusion rate of a slug of
tracer into a batch mixer can be characterized by a single quantity, a decay rate constant. This constant is then used
to define a dimensionless mixing-rate number.
Experiments in turbine-agitated and propeller-agitated fully-baffled tanks verify tbe theory, and relate this
mixing-rate number to the stirrer Reynolds number and to the Power number for mixing. These new correlationships
allow calculation of time needed or energy input needed to achieve any desired degee of uniformity of the mixture.

There are two aspects to the design of stirred tanks: IMF'ULSETRACER EXPERtMEhTS IN BATCH hDXEXB.DECAY
impeller power requirement and quality of mixing. The RATEOFTIIEFLUCTUATINGRESPONSEASAMEASUREOF
first aspect has been studied extensively in terms of the MIXINGRATE
power correlation for marine type propellers[l] and for Mixing takes place by a complicated interaction of
different types of turbines[2-51, and is summarized three mechanisms: bulk flow, turbulente and molecular
elsewhere[&81. The second aspect to design, the quality diffusion, and observations, confirmed by experiment [26],
of mixing or the rate of approach to uniformity, is the show the genera1 simplified behavior of Fig. 1. The smal1
subject of this paper. For this purpose, the power per unit well-mixed volume near the stirrer guarantees random
volume has been used as the scale-up criterion[6,7]. redistribution of fluid elements into one-passage RTD,
Although the power estimation is simple and fairly f(t), allowing the application of convolution for subse-
accurate, it gives only a crude approximation to the quent flow passages, as shown in Fig. 2.
quality of mixing. Let US assume that we make an impulse injection of
More direct methods for determining mixing quality tracer near the stirrer tips and measure the response
have been developed by measuring concentrations within nearby. Thus, injection and measurement are made in the
mixing tanks [9-151. Most authors defined “mixing time” smal1 well-mixed volume. From the material balance, the
as that required to reach an arbitrary deviation from the transfer function representation of an impulse response
final concentration: thus 0.1% deviation [ 101, 5% Y0) is t251
deviation[ 141, or visual observations [11-131. The first
mixing time correlations were made by Fox and Gex[ 121
for propeller and jet mixing, and by Norwood and
Y(s)
-=-=
T
F(s)
l-F(s) cDO
=,
Fk@)
.
Metzner[l3] for turbine mixing. Since then, mixing time
correlations based on these criteria have been widely used Note that this representation is valid only when fluid
for scale-up. However, the terminology has been defined elements are randomly redistributed into f(t) and that
differently by different authors, and thus the correlations y(t) is normalized to have a final value of one. By taking
are not useful for quality of mixing different from that the inverse Laplace transformation, eqn (1) becomes
defined (0.1% or 5% deviation from uniformity). A
criterion using the rate of decrease of concentration
fluctuations[l5] would be useful, but this approach has
not been wel1 studied. Mixing rates can be related to
recycling models, however this has been only vaguely
(2)
touched upon in the literature, even though there are
many studies on mixing behavior in terms of recycle where
models [ló-261.
The purpose of this study is to establish the relationship
between an experimentally tested recycle mixing mode1
j”(f) = j(l)*jct>n..
.“j(t)’
and the rate of approach to uniformity in the mixing tank, and the symbol * stands for convolution. Equation (2)
to develop a new and genera1 measure for this rate of shows that the impulse response is the sum of the
mixing, to compare its usefulness to the presently contributions of subsequent flow passages.
accepted measures, and to present correlations for finding
the mixing time or power requirement in turbine and Tanks in series with recycle
propeller agitated vessels, based on experiments (342 As the simplest representation, consider the equal-size
runs) in two different sized tanks (dia = 0.56 m, 1.22m). tanks-in-series model for the one-passage RTD, j(t). The
570 S. J. KHANGand 0. LEVENSPIEL

Roots of the denominator of the above equation are

l
N Lazy movement with
sk=-[l-expbj_N)]$ k=0,1,2 ,....
little mixing

_-- Welt mixed

After evaluating the coefficients of partial fractions and
taking inverse transforms, y(t) becomes

(al Turbine agitated mixing tank.

+Zgexp[-N(l-cos$)$]

cos (5)

where
-- Lazy movementwlth
little mixlng
B = 0; N =odd
1; N=even
.-- Wel1 mixed

M= (N-1)/2; N=odd
N/2-1; N=even’

Equations (4) and (5) are identical since both equations

represent the same impulse response. However, eqn (5) is
(bl Propeller agltated mlxing tank. useful for large t because when t is large and 2n/N < 1,
Fig. 1. Genera1flow pattern of fluidin mixingtanks. eqn (5) reduces to the approximation

Impulse
injection One di3i~lonal measurement , y(t) y(f)-1+2exp(-gf)cos($f+$) (6a)

-n- &-, P-; with decaying amplitude A

f(u)

mean-T
variante - u2 (6b)

Figure 3 shows an example of the impulse response

Fig. 2. Recycle model for a batch stirred tank. generated by the tanks-in-series-with-recycle model. As
can be seen, the approximate solution and its decaying
transfer function for this model is amplitudes (broken lines) given by eqns (6a) and (6b)
fairly wel1 represent the exact solution (solid line) given
by eqns (4) or (5).
F(s)= l N

( -Ts+l
N > Approximation for the genera1 recycle equations for an
arbitrary one-passage RTD using the centra1 limit theorem
where N is the number of tanks. Inverting eqn (3) and The above model is rather restrictive, so let USrelax the
substituting into eqn (2), the impulse response for a assumption that requires flow to follow the tanks-in-series
recycle system becomes model. Let USinstead represent the circulation of fluid in
the mixing tank with a reasonably narrow, but arbitrary
one-passage RTD, with mean T and variante u*.
Y(f) = 2 N(Nt’T)“-’
k=,(kN l)!
exp
-
(- Nt/T) (4) The centra1 limit theorem in statistics is a useful tool for
dealing with this kind of distribution and for relating the
Although this equation has been frequently cited in the distribution to the normal distribution. Feller [27] gives a
literature, it is difficult to use when t is large. We therefore complete discussion and proof of the centra1 limit
change it into a more useful form, first by substituting eqn theorem, however for our purpose, we simply need to
(3) into eqn (1) to give note that the tracer distribution after n passes approaches
the normal distribution as n +m. or
Y(s)= ,m T\N
(Gs+l) -1 f”W=&exp(-5); forlargen. (7)
 New scale-up and design method for stirrer agitated batch mixing vessels 571

=
.._
to aPPrOXimaie Solution. eqn(6b)
%.

“~,~~~~~~ __

,,,’
0
J
” I 2 3 4 5 6
t/T

Fig. 3. An example of the impulse response generated by the tanks-in-series-with-recycle model for a batch rnixing
tank.

Proof of this special form of centra1 limit theorem has where K is 2?r2u0*/T.Equations (8) and (9) are good for
been presented by Feller [28] and Paynter [29]. Recently any f(t) with smal1 ot. The decaying amplitude A of eqn
Mann et al. [30] recognized this in recycle systems. With (9) is of particular interest for mixing tanks since it is a
this theorem we need only two parameters, the mean and measure of maximum concentration deviation in the main
variante of f(t), to describe y(t) as it approaches its final flow Channel.
value. To demonstrate the generality applicability of eqns (8)
To determine how any arbitrarily narrow f(t) ap- and (9) with other types of single passage RTD curves,
proaches its final value, we can assume any reasonable consider the recycle model shown in Fig. 4. Figure 5
one-passage RTD with small variante. Tanks of uniform shows a typical impulse response of this model. As y(t)
size in series is a reasonable choice since a simple approaches 1, eqns (8) and (9) are good approximation for
analytical solution is available for large t and hr, thus the y(t) and its amplitudes. For different values of N and a of
dimensionless variante of f(t) is Fig. 4, amplitudes are calculated numerically and plotted
on semi-log scale for a comparison with the prediction of
,-,ez
= a’/ TZ = $. Plug flow Nequal-size tanks
AiW

1 @J$&& _... -Q-- f

Substituting into eqn (6) gives the impulse response -1
msan= T, mean=T2

a= q/T2

T=T,+T,

with decaying amplitude

Az2exp(-qr)=2exp(-Kt) (9) Fig. 4. An example of recycle models to demonstrate the validity

of eqns (8) and (9).

3 I\ l
N=2
a 92.5355
$90.04

Decoying amplitude
to approxlmate solutlon. eqn(9)

C
Approximate solution, eqn(6)
T

,’
O-
0 I 2 3 4 5 6

Fig. 5. An example of the impulse response generated by the recycle model shownin Fig. 4.
512 S. J. KHANGand 0. LEVENSPIEL

Four symmetrlcolly
- $aacd b&les,
.

d
E
a
f

-1
0.1

D--
1
Fig.7. Tank geometry and impeller position.

Injection diffuser
_--*-. / \

-- 03

Fig. 6. Semi-log plot of amplitudes for the recycle model shown in

Fig. 4.
Na ’
+ 2 0.5 0%2
? 2 2.0 0.056
x 5 0.5 0.086
x 5 1.0 0.05
A 10 0 0.1
A 10 0.5 0.044
a 10 2.0 0.011
0 25 0 0.04 i-_ 1.2d 4
I 50 0 0.02
(a) Turbine (b)Propeller

eqn (9). As shown in Fig. 6, eqn (9) wel1 represents the Fig.8. Positionof injectiondiffusersand conductanceprobes.
amplitudes for distributions with smal1 variances.
Thus we conclude that eqn (6) reasonably approximates wave injection unit which gave a 0.5 s tracer pulse in the
the response of the recycle system of Fig. 2 having an large tank, and 0.25 s in the smal1 tank. Since this was
arbitrary but not too broad single passage RTD. much smaller than the mean of the one-passage RTD, it
was reasonable to assume an impulse injection.
EXPEFUMENTAL
MEASUREMENTS
OF AMPLITUDES The four conductance probes were connected in
OF IMPULSE RESPONSES IN TURBINE- parallel to a conductance measuring circuit, thus giving
AND PROPELLER-AGITATED FULLY-BAFFLED TANKS the average concentration reading of the probes. In
Experimental setup addition, a fifth probe was attached to an injection diffuser
Two sizes of cylindrical flat bottomed tank (D = to detect the abrupt rise in concentration during injection
1.219m, 0.559 m) and two types of agitator (propeller and and start the recorder.
turbine) were used in this study; see Fig. 7 for the tank The probes and electronic circuit are unique in design
and baffle geometry, and impeller position. Propellers with the following features:
were three-blade marine type with pitch:diameter of
(i) linearity in output signal with respect to conduc-
1.5: 1. Turbines were standard six-blade disk type with
tante,
diameter: blade length : blade height of 20 : 5 : 4. For each
(ii) insensitivity to stray currents,
tank, two sizes of propeller and turbine were used.
(iii) high resolution and,
City water was used for the low viscosity runs and 70%
(iv) Y axis leve1 adjustable to cancel out background
glycerol for the high viscosity rum. The conductivity
concentration level.
method with strong electrolyte was judged to be most
convenient for this type of study so a saturated solution of Thus a whole series of tracer runs were made without
NaCl (1.5cm3 for the large tank, 5 cm’ for the smal1 tank) recalibration even though the average concentration leve1
served as tracer. in the tank rose after each run. For details of construction
Four porous glass injection diffusers and four conduc- of the probes and the electronic circuits, see Khang and
tante probes were positioned in the main flow Channel as Fitzgerald[32]; of the whole experimental setup see
shown in Fig. 8. The diffusers were connected to a square Khang[33].
 New scale-up and design method for stirrer agitated batch mixing vessels 573

Amplitude decay rate constant and its correlation with

stirrer Reynolds number
Table 1 summarizes the experimental condition used in
this study, and Fig. 9 illustrates typical impulse responses
for turbine and propeller agitators. In general, turbines
give a smoother flow pattem with less noise than do
propellers (Fig. 9a VS 9b), large tanks have a smoother
response than smal1 tanks (Fig. 9c VS 9a); and high
viscosity liquids are more noisy than low viscosity liquids
(Fig. 9d VS 9a) probably because of insufficient mixing
near stirrer tips.
For a particular stirrer, tank, and fixed stirrer rpm,
zOTlme, sec &lO impulse responses were recorded. These were
Ia) normalized by dividing by their final values, and their
-1 peak and valley values were recorded, except for the first
few. The first few were avoided in order to minimize
effects from abnormal shapes of the one-passage RTD.
The peak and valley values were changed into the
amplitudes from the final value, A, and then plotted VS
time.
From these A VS t measurements, we can find the
decay rate constant K by taking logarithms of both sides
of eqn (9), or

1nA =ln2-Kt where K =2r2ue2/T.

40
Time. sec Thus from theory the slope of the In A VSt plot gives K
(bl while the intercept should be at In 2. Figures 10 and 11
show typical 1nA VS t plots for turbine and propeller
agitators. After forcing the line through the theoretical
intercept of In 2 least square linear regression was used to
evaluate the slopes.
From dimensional analysis, we can show that the decay
rate constant K is related to the system variables by the

0 20 40

T'm% sec

-3 ~

10 20
Time >sec
(d)
-4rt.
Fig. 9. Typical impulse responses in mixing tanks uuder various 0 5 10 15 20
conditions. (a) Run A31; D = 0.559m, d = 0.127m, n = 160min-‘,
1. sec
water 17.5”C.(b)RunG19; D = 0.559m, d = 0.114m, n = 94 min-‘,
water 19°C.(c) Run E29; D = 1.219m, d = 0.366m, n = 80 min-‘, Fig. 10. A typical In A VS t plot for turbine. This graph represents
water 10°C.(d) Run D26; D = 0.559m, d = 0.224m, n = 90 min-‘, al1 10 runs at 80 rpm. D = 1.219m, d =0.366m, n = 80min-‘,
p’= 1.07X10~‘m’/s. water 10°C.
574 S. J. KHANG and 0. LEVENSPIEL

Table 1. Summary of experimental conditions

high Reynolds number, the functional form in eqn (10) is a

-1 Intercept = In 2
Propeller
1 constant. Thus, for turbine mixers,

~0.5; Re>2X 103 (11)

1
while for propeller mixers,

__.~ = 0.9; Re > 10’.

-1 (12)

Now from accepted correlation of Power number with

--__ -i Reynolds number [2,5,6], the relationship between
Mixing-rate number and Power number is, for turbine
mixers,

n d 2*3~&~O~5;Re>l(f
-I KD >(> pn d
(13)

while for propeller mixers,

-l n d 2z15-=0.9;
Pa
20 40 60

Fig. 11. A typical In A VS t plot for propeller. This graph

80
KD >(>
’ pn3ds-
Re> 10”. (14)

represents al1 9 runs at 90 rpm. D = 1.219m, d = 0.254m, These equations relate time of mixing, power requirement,
n = 90 min-’ , water 10°C.0,2nd peak; @,2nd valley;8,3rd peak. stirrer size and RPM needed to achieve any required
degree of uniformity for a given fluid, and are useful for
dimensionless groups scale-up.
The following example illustrates one of the many types
(;)(_t!>’=f(+> (10) of problems which may be treated with theserelationships.

Example
where n, p, p, d and D are the rotational speed, density, A smal1 slug of foreign fluid is added to a large fully
viscosity, diameter of stirrer and diameter of tank. The baffled tank of water (D = 6m, p = 1000kg/m3, /L =
dimensionless group n/K represents the rate with which 0.01 kg/m.s) containing a centrally placed 6 blade disk
the mixture approaches uniformity, and we name it the type turbine (d = 0.6 m). What horsepower motor and
“mixing-rate number,” Nti,. Thus the above equation can RPM are needed to insure that concentration fluctuations
be rewritten be less than ? 0.1% after 5 min of agitation?

0$ ’ =
Nmix f(Nac).
Solution
First find the stirrer rotational speed for a 5 min mixing
Figures 12 and 13 show the relationship of eqn (10) for time. From eqn (9),
turbine and propeller mixing. The exponent c on the
geometrie factor in eqn (10) was estimated for best fit. The K = In
-_ (Alz) _ In (O.~lD) = 0 02534 s-~
value found was 2.3 for turbine and 2.0 for propeller. For -t -300 . .
 New scale-up and design method for stirrer agitated batch mixing vessels 575

0.75. Turbine ---

: (six flat blades. disk type)
\
.

\
: 0.5. l
\. J._J__ . __-.--___..
ô3
x . . 0 0 . o \ 0.406
h
S
0.25..

IOruns/point

0 ‘, 1
103 10' 100
Re(=rdp/p)
Fig. 12. Mixing number group VSReynolds number for turbine-agitated batch mixers.Each point represents the data
in a graph such as Fig. 10.
D(m) d(m)
0 1.219 0.366
0 1.219 0.488
0 0.559 0.127
IO.559 0.244

Propeller
(marine type,
ISpitch. three blades) -c--
6 - IOruns/point
I

Fig. 13. Mixing number group VSReynolds number for propeller-agitated batch mixers. Each point represents the
data in a graph such as Fig. ll.
D(m) d(m) *
0 1.219 0.114
0 1.219 0.254
0 0.559 0.114
n 0.559 0.254

Assuming that the Reynolds number is larger than lO”,the Next, the required power is, from eqn (13),
stirrer rotational rate is calculated from eqn (11) or (13) to
be P = 0.5 pn’d’/g, = (0.5)(1000)(2.53)3(0.6)5

n = 0.5 K(D/d)‘.’ = (0.5)(0.02534)(6/0.6)2’9 =628W (0.84 Hp).

= 2.53 ss’ (152rpm). Therefore, a motor with the minimum specification of

152rpm and 0.84 Hp is needed to achieve the desired
To check the assumption, Re > lO’, degree of mixing in 5 min.

Re = (2.53)(0.6)‘(lOOO)/(O.Ol)
= 9 x 10’ Comparison with previous investigators’ correlations
Table 2 summarizes previous investigators’ experimen-
thus, the assumption is justified. tal conditions, correlations and mixing times. Tank and
576 S. J. KHANG and 0. LEVENSPIEL

Table 2. Comparison with previous investigators’ correlations

Mixing
Definition Reported time,
Mixing time of Mixing calculated
correlation Mixing time time from Eq. 9
* Last

nt.0.905(3 oepen& a* 373.8 5 299.4 5

A for
.log(Z/A)'
A=O.OOl
for Re ) UI4
Last
(nt)($1'5 indicator
350.7 s ----
n2d 1,6 ($1
z *= 3o :;a:;e
(TI
for Re=3.2E4 i 13.4 A=O.O5 156.6 5 184.4 5

"r=3.48(%2.05 Depends on 644.5 5 380.0 5

d
A for
-log(zlnl*
for Re , 104 A=O.OOl

* ~ommon logarithms

baffle geometries are the same as those of Example 1. The Prochazka and Landau [ 151injected tracer solutions at
last column gives the mixing time calculated by eqn (9) the liquid surface in mixing tanks and experience shows
and Figs. 12 and 13 for comparison, even though some that molecules at the surface tend to linger there.
experimental conditions are different from those of Figs. Therefore, their correlations give considerably longer
12 and 13. mixing time than does eqn (9). However, the exponents on
Norwood and Metzner[l3] used the method of acid- the geometrie factor in their correlations (2.57 for turbine
base indicator and determined the mixing time by a visual and 2.05 for propeller) are close to those in eqn (10) (2.3
observation of the last indicator color change. Their for turbine and 2.0 for propeller).
mixing time, 130.7s, for a turbine is considerably shorter As shown above, the present mixing rate correlation
than a 0.1% deviation time, 299.4 s, calculated by eqn (9) given by eqn (9), Figs. 12 and 13 roughly checks previous
probably because the visual observation of indicator color correlations, within the limits of experimental conditions.
change was not as sensitive to concentration variations as
a conductivity probe. In fact by back-calculation, we CONCLUSIONS
estimate that the limit of detection of color change in their We draw the following conclusions from this study on
experiment is at about A ~7%. stirrer agitated fully-baffled mixing tanks.
The mixing time correlation for a propeller by Fox and (a) With RTD theory and probability conditions (cen-
Gex[l2] was also based on the method of acid-base tral limit theorem), we show how a slug of foreign fluid
indicator. Their mixing time, 350.7 s, would have been blends into the vessel fluid. Equations (8) and (9) tel1 how
much shorter had they used fully-baffled tanks. uniformity is approached.
The correlation by Kramers et al. [ 101gives the mixing (b) This prediction is verified by experiments
time of the same order but somewhat smaller than that (342runs).
predicted by eqn (9). This may have the following cause. (c) As a result of this theory, experimental results of
They placed two electrolyte conductivity probes in a Figs. 12 and 13, and Power number Reynolds number
mixing tank, one near the surface, the other near the correlations of the literature, we develop a design method
bottom of the tank. They measured concentration to relate mixing time needed and mixing power require-
differences between these two probes instead of measur- ment to achieve any degree of uniformity for any physical
ing concentrations at each probe separately. Since the system and mixer geometry.
flow pattern and the concentration distribution of the (d) Example 1 shows how to use these equations.
upper region of a turbine mixing tank tends to mirror the
lower region of the tank, the concentration readings at the Acknowledgements-The authors thank Mixing Equipment Com-
two probes tend to follow each other, thus concentration pany for the loan of their mixers and to Crown Zellerbach
differences between the two probes would be smaller than Corporation for the laan of their equipment. This work was done
as part of National Science Foundation Grant GK37151and one of
concentration variations of each probe. Thus, their mixing the authors (S.J.K.) thanks N.S.F. for the research assistantship
time is shorter than that of eqn (9). provided by this grant.
Biggs [ 141 used one electrolyte conductivity probe to
measure concentration fluctuations. His mixing times for NOTATION
turbine and propeller are a little shorter than those of eqn a (mean residence time for the plug flow
(9) probably because a smal1 amount of through-flow was region)/(mean residence time for the tanks in
maintained and tracer injections were made at the inlet series regions), dimensionless, see Fig. 4.
pipe. A normalized amplitude of response signal, or
 New scale-up and design method fol. stirrer agitated batch mixing vessels 511

(y - ysna,)envelope/(y~nal
- yinitdt dimensionless, [6] BatesR. L., Fondy P. L. and Fenic J. G., Mixing (Edited by V.
see Fig. 3. W. Uhl and J. B. Gray), Vol. 1. Academie Press, New York
d diameter of stirrer, m, see Fig. 7.
[7] I!IEnd F. A. and Chapman F. S., Liquid Mixing nnd
diameter of tank, m, see Fig. 7. Processing in Stirred Tanks. Reinhold, New York 1966.
rR one-passage RTD, or single circulation RTD, SC’ [8] Sterbacek Z. and Tausk P., Mixing in the Chemical Industry,
F(s) Laplace transform of f(t), dimensionless International Series of Monograph in Chemical Engineering,
Newton’s-law conversion factor, unity and dimen- Vol. 5. Pergamon Press, London 1965.
&
[9] Fossett H., Trans. Instn. Chem. Engrs. 195129 322.
siordess for SI units [ 101Kramers H., Baars, G. M. and Knoll W. H., Chem. Engng Sci.
g-1, dimensionless 19532 35.
L amplitude decay rate constant, SC’ [ll] Van De Vusse J. G., Chem. Engng Sci. 19554 178.
n stirrer rotational speed, SC’ [12] Fox E. A. and Gex V. E., A.LCh.E. J. 1956 2 539.
[13] NorwoodK.W.andMetznerA.B.,A.I.Ch.E.J.1%06432.
N ln,n n/K, mixing-rate number, dimensionless [14] Biggs R. D., A.I.Ch.E.J. 1963 9 636.
P mixing power requirement, W [IS] Prochazka J. and Landau J., Colln. Czechoslou. Chem.
Re nd’p/p, stirrer Reynolds number, dimensionless Commun. 196126 2961.
RTD residence time distribution, SC’ [16] Van De Vusse J. G., Chem. Engng Sci. 196217 507.
[17] Van De Vusse J. G. Chem. Engng Sci. 196621 611.
t time, s
[lg] Holmes D. B., Voncken R. M. and Decker J. A, Chem. Engng
T mean of f(t), s Sci. 196419 201.
YU) impulse response of a recycle system, dimension- [19] Voncken R. M., Holmes D. B. and Den Hartog H. W., Chem.
less Engng Sci. 196419 209.
Laplace transform of y(t), s [20] Clegg G. T. and Coates R., Chem. Engng Sci. 1967 22 1177.
Y(s) [21] BuffhamB.A.andGibilaroL.G.,A.I.Ch.E.J. 1%814805.
ZL. liquid depth, m [22] Gibilaro L. G., Chem. Engng Sci. 1971 26 299.
[23] Levenspiel O., Chemical Reaction Engineering, 2nd Edn.
Greek symbols
Wiley, New York 1972.
[24] Hochman F. A. and McCord J. R., Chem. Engng Sci. 1970 25
u* variante of f(t), s2 97.
oa’ <r’/T*, dimensionless variante of f(t) [25] Fu B., Weinstein H., Bernstein B. and Shaffer A. B., I& EC
P density, kg/m3 Proc. Des. Deo. 197110 501.
CL viscosity, g/m.s
[26] Reith Ir. T., A.LCh.E.-I. Chem. E. Symposium Series 1965 10
15.
[27] Feller W., An Introduction to Probability Theory and Its
Applications, Vol. 1. 3rd Edn, p. 243-261.Wiley, New York
RRFERENCES 1968.
UI Rushton J. H., Costich E. W. and Everett H. J., Chem. Engng [28] Feller W., An Introduction to Probability Theory and Its
Progr. 1950 46 395. Applications, Vol. 11.p. 487-491. Wiley, New York 1%6.
PI Bates R. L., Fondy P. L, and Corpstein R. R., I& EC Proc. [29] Paynter H. M. Regelungstechnik. Heidelberg Conference
Des. Deu. 19632 310. Proceedings, Oldenbourg, Munich 19564 243.
[31 Olney R. B. and Carlson G. J., Chem. Engng Progr. 1947 43 [30] Mann U., Crosby E. J. and Rubinovitch M., Chem. Engng
473. Sci. 197429 761.
[41 O’Connell F. P. and Mack D. E., Chem. Engng Progr. 1950 46 [31] Hansen P. D., Simulation 1966 6 322.
358. [32] Khang S.J. andFitzgeraldT. J., ILECFundl, 197514208.
Vl Rushton J. H., Costich E. W. and Everett H. J., Chem. Engng 1331
_ _ Khanr!S. J.. Ph.D. Thesis, Oregon State University, Corvallis,
Progr. 195046 467. Oregon 1975.